We consider droplets of radius R injected in a porous medium with a pore si
ze d less than or equal to R. The droplets react chemically with the surfac
e of the pores, and make it less wettable. They should acquire a velocity V
which is, in most cases, independent of R. For R similar to d, we expect t
he trajectory to be a random walk. For R > > d, however, it should be balli
stic. These effects should be visible when gravity effects are negligible:
either by using very small pores (d < d* similar to 10 mu m), or saturating
the porous medium with a passive fluid of matched density. Various transpo
rt problems associated with this process are discussed here at the level of
scaling laws. (C) Academie des sciences/Elsevier, Paris.